Several measurement techniques are available for determination of phase transition points of pure substances, blends, and solutions which can be selected according to experimental conditions and other specific requirements. The main issue of all below discussed techniques is that they try to determine an equilibrium property (a freezing/melting point) during heating or cooling a sample, where the system is not in an equilibrium state.
Determination of invariant points (degree of freedom f = 0) in single-component systems (triple points), can be performed comparably easy and with high accuracy (better than 1 mK). Therefore, several reference values of the international temperature scale are based on these points, such as the triple point of water at 273.160 K. In single-component systems the determination of univariant points (f = 1) is much more complex but often possible with high accuracy because they are reduced to invariant points when the determination is performed at constant pressure. This was shown for acetonitrile [100-102], a single-component system with very little tendency to supercooling. [103] Even at constant pressure x(T)-lines of binary solid-liquid phase diagrams are generally not easily determined as - for eutectic systems - the composition of the phases changes during heating or cooling (f = 1).
Examination of melting and freezing points for preparation of such phase diagrams is relatively often performed by application of differential thermal analysis (DTA) and differential scanning calorimetry (DSC) [104-106]. The advantages of both methods are their fast and comparatively easy practicability and the very small sample volumes that are required; the latter is of major interest when examining very expensive substances. The main disadvantages are however a lower accuracy [100] and higher sensitivity to systematic errors. The reduced accuracy is mainly caused by the small sample volumes because in this special case already smallest impurities can strongly falsify the phase transition temperature by means of a lowered melting point and raised boiling point as well. Additional effects that are caused by the wall of the sample vessel become strongly apparent by examination of such small sample volumes. [103] Another factor that is responsible for the reduced accuracy of these methods is generally the lack of a stirring device. Hence, mass transport and heat transfer are exclusively performed by comparatively slow diffusion and weakly pronounced convection that leads to anisotropic temperature distribution and, in the case of mixtures, to an inhomogeneous sample composition. [103] The latter is a major disadvantage for determination of phase transition
points in multicomponent systems, where a homogeneous sample constitution is a crucial point. The anisotropic temperature distribution, combined with generally applied high cooling rates, enhances the supercooling of studied samples. [107,108] Supercooling is an important error source during investigations of liquid to solid phase transitions. For common organic solvents supercooling can reach a level of about 15 K, but for ILs the supercooling effect is even worse and can reach up to 200 K [109] which strongly falsifies the determined crystallisation points. The large supercooling, in combination with the generally applied high cooling and heating rates, made a determination of crystallisation and melting points for some ILs impossible because the samples do not crystallise within the working range of common DSC and DTA devices.
An alternative to the above discussed methods is the investigation of cooling curves proposed by Andrew, Kohman and Johnston [110]. Additional examples of this method are shown in Refs. [100,111]. This method requires observation of the temperature of a sample which is slowly cooled down or heated up. Using this procedure, greater accuracy is achieved, because much larger probe volumes are used. Small cooling and heating rates can be applied to reduce supercooling, because they have no negative influence on the sensitivity of this method. [103] As shown by Rossini et al. [100] the melting point of pure substances, not tending to supercooling, can be measured with an uncertainty of ± 10 mK.
A disadvantage of this method, when compared to DSC and DTA, is the lower sensitivity, which makes determination of phase transitions, where only little energy quantities are transferred, nearly impossible.
Another method is described by Schrödle et al [112]. This method is similar to the cooling curve method but the phase transition is determined with a photo detector instead of a temperature sensor and the temperature of the thermostat bath at this point is assumed to be the phase transition temperature. [103] The disadvantages of this method are its restriction to transparent samples, the assumption of identical temperatures of the sample and the thermostat bath and it gives no information if supercooling has occurred.
All of the above discussed techniques require exposing the sample to a temperature gradient with simultaneously recording of its temperature as a function of time. If a phase transition occurs in a single component system a halt is observed due to the released phase transition enthalpy that counterbalances the cooling process of the sample. In multi component systems the phase transition enthalpy results in a breaking point because only one component crystallises and the composition of the remaining phase changes with time.
Ideal examples of temperature-time curves of single-component and multicomponent systems respectively are shown in Figure 2-6.
A B C D E F
Cooling Curves Heating Curves
a b
c
d e f g
b a
g f
e d
c
T1 T2
T3
Figure 2-6: Examples of idealised cooling and heating curves of a pure substance and a binary mixture; according to Ref. [100].
In general, there are three types of cooling and heating curves which can be observed.
1) Crystallisation of a pure substance (curve A of Figure 2-6)
The temperature of the sample varies with the applied temperature gradient until the freezing point T1 is reached. This temperature is kept constant (a-b) by the phase transition enthalpy compensating the thermal loss to the surrounding. After the substance is completely crystallised, the temperature of the sample decreases continuously again according to the applied temperature gradient, following Newton’s equation.
2) Cooling of a binary mixture (curve B of Figure 2-6)
The temperature of the sample varies with the applied temperature gradient until point c at temperature T2 is reached. At this point, one substance starts to precipitate from the liquid. The released phase transition enthalpy compensates only part of the heat loss of the sample to the surroundings, resulting in a reduced slope of the cooling curve (breaking point). If the eutectic composition at T3 is reached, the liquid crystallizes completely without further variation of its composition, resulting in a eutectic halt (d-e).
3) Cooling of a eutectic mixture (curve C of Figure 2-6)
If the eutectic temperature T3 is reached, the system crystallizes completely forming a solid with eutectic composition.
The curves D, E, and F (Figure 2-6) show heating curves which are complementary to the cooling curves A, B, and C (Figure 2-6).
During a real cooling experiment delayed crystallisation or supercooling can occur, which leads to a curve like the one shown in Figure 2-7. Therefore, it can be necessary to determine the melting point by extrapolation of the horizontal or quasi-horizontal parts of the curves. The intersection point of the two curves determines the crystallisation point.
[103,107,108]
Figure 2-7: Nearly ideal cooling curve of a common organic solvent (γ-butyrolactone) determined at a cooling rate of -30 K h-1.
Unfortunately, supercooling in the case of pure ILs and their mixtures is too large to get viable freezing points, even with the extrapolation procedure mentioned above. Therefore, in the case of ILs, melting points were investigated instead of freezing points. The observation of melting processes with this method suffers some drawbacks just as their determination with DTA/DSC, because it is very complicated to sufficiently mix a sample which contains a large amount of solid. Additionally, a solid, which is in the sample container, melts from the edge of the sample container to the centre, where the temperature sensor is usually placed. This sensor is then still surrounded by the solid, despite an already started melting process. But despite these theoretically based drawbacks, melting points of organic solvents determined with this method are in good agreement with the
corresponding freezing points and values from literature as well, along with a high reproducibility [107,108]. Therefore, in this work melting points of ILs were determined, because they are assumed to be less affected with errors than the corresponding freezing points.
To support the determination of phase transition points and to verify the results obtained from cooling and heating curves, the conductivity of the sample was simultaneously recorded with its temperature. Realization and evaluation of the determination of phase transition points is explained in Chap. 3.4.